In partnership with
Moonshot AI has released Kimi K3, a 2.8 trillion parameter model the Chinese lab calls the largest open-weight system ever built. Announced in mid-July, it arrives with frontier level benchmark claims and a full weights release scheduled for 27 July.
The significance runs deeper than one launch. A Chinese open model priced aggressively and closing on the best US systems reshapes the frontier conversation and revives hard questions about how open weights get governed and secured.
What can it do?
2.8 trillion parameters, billed as the largest open-weight model yet
Roughly 1 million token context window with native vision support
Weights due 27 July on Hugging Face, free to self-host
Claims to top Claude Fable 5 and GPT-5.6 Sol on some coding tests
API priced near $3 per million input, $15 per million output tokens
Your prompts are leaving out 80% of what you're thinking.
When you type a prompt, you summarize. When you speak one, you explain. Wispr Flow captures your full reasoning — constraints, edge cases, examples, tone — and turns it into clean, structured text you paste into ChatGPT, Claude, or any AI tool. The difference shows up immediately. More context in, fewer follow-ups out.
89% of messages sent with zero edits. Used by teams at OpenAI, Vercel, and Clay. Try Wispr Flow free — works on Mac, Windows, and iPhone.
Why does it matter?
Open weights let any developer or government run a near-frontier model locally, sidestepping US API providers and their pricing.
The benchmark claims are vendor reported until the weights land on 27 July, so independent testing will decide whether the hype holds.
Our take
The pattern is familiar. A year on from DeepSeek's market moving debut, another Chinese lab has shipped a competitive open model at a fraction of Western prices. China's bet is that diffusion beats secrecy, with good-enough intelligence spread widely driving economic productivity faster than a walled superintelligence. Kimi K3 is that thesis at 2.8 trillion parameters.
For marketers and builders the takeaway is leverage. When a frontier-class model is free to download, the cost of experimentation collapses and subscription tools face real pricing pressure. The open question is governance. Weights anyone can run are weights nobody can recall, and the safety debate gets louder the moment K3 reaches Hugging Face on 27 July.
And another big thing… Apple takes its OpenAI fight public
Apple has sued OpenAI and two former employees, alleging they lifted confidential hardware trade secrets to fuel OpenAI's device ambitions with the Jony Ive founded startup io. OpenAI rejects the claim, saying it has "no interest in other companies' trade secrets."
The dispute sharpens a widening Silicon Valley talent war and puts OpenAI's credibility under scrutiny as it weighs a future public offering. A courtroom skirmish over a metal-finishing technique has become a strategic business risk.
